A few months ago, I decided to see just how fast I could generate images with Stable Diffusion. I used ComfyUI and Stable Diffusion XL Turbo, attempting to make an “Animorphs cover” capability with a simple slider. This isn’t exactly how diffusion works, but the result wasn’t too bad!
More recently, I saw someone post a two-axis chart. He would drag a dot around the chart, changing the image being generated. His setup pre-calculated all of the images, and this would only change which one was displayed.
I knew that I had to replicate this, but with near realtime generation. While I was at it, I decided to make four variables instead of just two. This would be way too many to pre-generate.
This was all on my personal PC with a gaming graphics card, the Nvidia GTX 3060 12GB. If you want to do this yourself,
- Get ComfyUI through https://pinokio.computer
- Download Stable Diffusion XL Turbo from Huggingface
- Drag this image into ComfyUI (it will load in the workflow):